Explore freely
Ask questions anywhere in the governed book. Inquiry is read-only: asking a question cannot change a model, estimate, or record.
Work the way you think.
The AI harness for fundamental research analysts.
Explore ideas with frontier AI. Test them through the actual models in a safe Sandbox. Preserve the work worth remembering.
Case Study: Where does AI-related debt exposure sit?
Inspect exactly
Before a Sandbox Run, see every proposed assumption against its current model value. Afterward, compare the result with a paired control on the same Version.
Preserve deliberately
A Run changes nothing in the governed baseline. Save the assumptions, results or conclusion only when they are worth remembering.
Inspect exactly / IRIS reads the workbook
Start with the model you already have.
Import the Excel workbook. IRIS identifies the historical periods, forecast periods, operating drivers, assumptions, and financial statements—and explains how they fit together.
IRIS works out the spreadsheet once—so every question after that starts with the business, not with the grid.
How the model works
Follow the drivers through the financial statements.
See how volume and pricing build revenue, how margins turn sales into earnings, and how investment and working capital affect cash flow and financing needs.
IRIS reads the formulas and assumptions already in the workbook. Its explanation gives you a starting point for asking what matters and where to look more closely.
Understand the forecast
Know what sits behind each estimate.
What drives the number? Which assumptions matter most? What evidence supports the view, and where is the model thin?
Follow an estimate back to its drivers and supporting research. That understanding stays with the model as the forecasts change, instead of being re-derived every time someone asks.
Preserve deliberately / Your Method, inside Excel
Build your approach into every estimate.
Every important estimate reflects an approach: which evidence matters, which assumptions hold, and what should cause the forecast to move.
IRIS makes that approach explicit, saves it as a Method, and puts it into Excel with =AI(). The estimate remains part of the workbook and flows through every calculation that depends on it.
=AI()
A Method is your approach to the estimate, running inside the model.
Draft it with AI or write it yourself. It shapes the estimate only after you approve it.
A Run executes the Method you approved—not whatever an assistant would say today. Revise a Method and you create a new version of it; replaying an earlier Run replays the Method that Run actually used.
Explore freely / Beside the workbook
Research the company beside the forecast.
Ask what supports an assumption, what changed in the latest filing, where the model is thin, or what would change the investment view.
What supports this margin assumption?
What does the latest filing say about funding needs?
Which part of the forecast looks least supported?
What would make me change this estimate?
IRIS answers with the model already understood. Ask as much as you like—asking changes nothing. A conclusion worth keeping goes back to the company and the estimate when you say so, instead of disappearing into another chat.
Inspect exactly / The governed Sandbox
Test a different view.
Change a market assumption or challenge the company outlook. IRIS proposes the specific cells it would change, and nothing runs until you approve that list.
Ask, “What happens if SOFR rises 100 basis points?”
Before the Run
Every proposed assumption, against its current model value.
IRIS finds where SOFR enters the forecast and shows each cell it proposes to change beside what the model says today. You approve a specific list of assumptions, not a description of a scenario.
Market values are frozen at the moment of the proposal. If the world moves before you approve, IRIS refuses the stale value rather than quietly running on it.
After the Run
Compared with a paired control on the same Version.
Pressing Run executes two Runs: a control carrying no assumptions, and your Sandbox bound to it—same Version, same Methods, same frozen market snapshot, differing only by the assumptions you approved.
The comparison is same-quarter, Sandbox against control. What you read is the difference your assumptions made, not movement that was already in the forecast.
A Run changes nothing in the governed baseline. Your working model, the current version, and every Method are exactly as they were. The Sandbox and its control are each preserved as immutable Runs, available for comparison—and a note is written only when you save one.
Preserve deliberately / Overnight analysis
Let the models keep working.
IRIS can refresh approved market and economic inputs overnight, identify which estimates depend on what changed, and rerun the affected analysis—using the Methods you approved, not a fresh opinion.
Start with what changed.
What moved?
Which forecasts held?
Which companies deserve attention?
What changed in earnings, cash flow, or financing?
Why?
The results stay available for review, with the assumptions, evidence, and prior work behind them. Begin the day knowing where to focus.
Explore freely / Across the book
See the whole book.
Turn your models into an ecosystem.
The same move in rates, spreads, demand, or pricing can affect every company differently. IRIS lets the PM ask one question across the models and follow every answer back to the relevant forecast, assumptions, evidence, and prior work.
PM View reads. It never writes. It can tell you what every model in the book says and show you exactly where each answer came from—but it never becomes the answer itself. The model, and the analyst who owns it, stay the authority.
When the world changes, which of my models should change with it—and why?
A move in credit spreads may matter to funding cost in one company, reinvestment yield in another, and expected losses somewhere else.
Compare how the same economic change affects different companies, using the assumptions and forecasts in each model.
Research in one name can inform the next. Compare exposures, test a shared scenario, and find the evidence behind a view across the companies you cover.
Where are higher rates hurting earnings?
Which companies need refinancing?
Where would wider spreads improve returns?
Which estimates depend on assumptions that have just moved?
The team's research becomes knowledge the whole book can use.
Preserve deliberately / Keep the history
Know why the numbers changed.
A revised estimate should not erase the view that came before it. IRIS keeps the approach behind the estimate, evidence, model change, and resulting financial impact together.
Go back to any version and see what changed, why it changed, who approved it, and how the change flowed through the model.
This is not chatbot memory. Nothing is absorbed in the background, and nothing is learned from you without being written down. Entries are explicit, reviewable, and versioned—and replaying an earlier run replays what that run actually used, not what the system knows today.
A worked inquiry / Explore, inspect, preserve
Case Study: Where does AI-related debt exposure sit?
Start with CoreWeave (CRWV). Trace the financing obligations, ask who bears the exposure, and examine what would change its value. Each answer sets up the next question, from the scheduled debt to the evidence needed for an investment view.
The inquiry starts with evidence already in the model: lease obligations, the cash interest rate forecast, delayed-draw term loan (DDTL) and OEM financing disclosures, undrawn borrowing capacity, and equity. From there, it asks whether the exposure can be connected to other companies in the book.
A useful finding from the inquiry: in the exchange below, IRIS identifies the gap between forecasting contractual debt payments and estimating a DDTL's market value. It identifies what is still needed: which loan is being valued and as of when, its expected cash flows, market pricing, contractual terms, and credit risks. The next question asks how to resolve each gap.
Follow the inquiry: 11 questions from obligations to an investment thesis
Where the debt sits
- Which financing obligations does this model schedule explicitly, and which does it only name from the filings without a row?
- For the delayed-draw term loans, what committed, drawn and undrawn amounts does the available evidence establish, and as of what date?
- How does the model treat the OEM and software financing, and is any of it double-counted with the disclosed debt total?
- What does the model assume about lease liabilities versus lease payments, and where would a committed GPU lease show up?
Who bears it
- For DDTL 3.0, who is borrower, guarantor, arranger, disclosed lender and known current holder, and which of those roles has a date attached?
- Which of those lenders is a public company in the book, and what would be needed to connect this facility to that lender's model?
How its value changes
- What drives the cash interest rate assumption, and what happens to interest expense if SOFR moves 100 bp?
- Which obligations reprice with market rates and which are fixed, and does the model separate them?
- What would have to be true about revenue or utilization for the scheduled repayments to be covered from operations rather than refinancing?
Where the model is thin
- What evidence is missing before this model could support a mark on the DDTL, and which of those items are in the filings already available with the models?
- Which of the unknowns you named would a saved note or a revised approach to the estimate resolve, and which need new evidence?
Carry the conclusion forward. Eleven questions, none of which changed the model. The analyst decides which of them produced something worth keeping, and saves that as a note with the model. The next time the financing question comes up, it starts from the evidence, the interpretation, and the questions that were still open.
The folder objection
Why not just point the model at the folder?
You can, and it will work better than most people expect. A frontier model will open the workbooks, follow the formulas, and tell you something plausible about them.
What the folder cannot tell it is which version was current, which estimate the analyst had already approved, what evidence that approval rested on, or what the last person to look at this company concluded. A folder holds files. It does not hold the state of the work.
IRIS holds that state. Use Claude or Codex to research the company, challenge an assumption, or reason through a forecast—against a book whose current version is resolved, whose approvals are explicit, and whose conclusions were written down on purpose. The research focuses on what drives the estimate: funding costs, unit growth, margins, credit losses, pricing, or operating leverage.
Your questions build on earlier research and the current forecast. You can see the assumptions behind a proposed change, decide whether to use it, and return to the reasoning later.
From one model to the whole book
Explore freely. Inspect exactly. Preserve deliberately.
Ask anything of the book. Nothing you ask can change it.
See every proposed assumption before it runs, and what it moved against a paired control.
Keep what you concluded, on purpose, where the next question will find it.
The analyst decides what becomes the record. IRIS makes sure the record survives the next question.
A frontier model can read a folder of spreadsheets. IRIS gives it the book instead—current, governed, and remembering what the analyst concluded. The models stay in Excel, the analyst decides the approach, and each new question builds on the work that came before it.